Adhere & V/TSIC: Derailment protection, mitigation and consequence estimation
At the RSSB sponsored Vehicle/Track System Interface Committee seminar, RSSB’s Dr David Griffin and University of Huddersfield’s Dr Philip Shackleton explored how the industry might explore additional measures for the guidance of derailed trains. This work was commissioned to follow up a recommendation from the RAIB investigation into the Carmont accident in August 2020 (recommendation 12 – see panel).
In the Carmont accident, a train derailed on debris washed from the side of a cutting. In 2016, a train derailed in a similar way just north of Watford tunnel.
The outcomes were very different and could have been a lot worse. Carmont happened during the pandemic emergency when there were few people on the train. Although three people sadly died, given the nature of the subsequent damage, a more crowded train would have resulted in a lot more casualties. At Watford, the train derailed into the six foot and was partially constrained when one of the running rails became sandwiched between the traction motors and/or gearboxes on the leading motor carriage. Although another train was approaching, the driver managed to send a code red alarm over the radio causing the driver of the approaching train to apply the emergency brakes before hitting the derailed train with a ‘glancing blow’ at comparatively low speed. These two examples highlight those two derailments, with a very similar cause, had quite different outcomes.
The research project, ‘Assessing the case for rolling stock and infrastructure design features that can provide guidance to trains when derailed’ (T1316) involves two strands.
Firstly, building on the comparison between Carmont and Watford, RSSB is developing a risk model to understand the benefits of derailment containment measures. It is assessing both location and route-specific derailment risks accounting for the features of the line of route, rolling stock, operational speeds, the operational environment, and passenger use. Using this work, the risk benefit from rolling stock and infrastructure upgrades can be assessed. It can also be used to provide cost benefit results for new lines, upgraded lines (renewals) and measures to address specific high-risk locations. This is a significant undertaking. The overall risk model combines four models: causal, trajectory, escalation, and loss.
The causal model calculates the probability of derailment in each 25-metre section, covering approximately 40 derailment causes. Probability is dependent on the assets present at the location (cuttings, level crossing, etc.) and train type (passenger, freight). It has been grouped in eight derailment types:
- Derailment on facing points (e.g., Potters Bar, Grayrigg).
- Derailment due to broken rail on a curve (e.g., Hatfield).
- Derailment at leading wheelset caused by striking a major obstruction (e.g., Ufton Nervet, Great Heck).
- Derailment of leading wheelset – not a major obstruction (e.g., Carmont, Watford).
- Non-leading wheelset – major rolling stock or track failure resulting in loss of support such as broken rail or failed bearing (e.g., Newton Abbott).
- Non-leading wheelset – Minor rolling stock failure or track failures where there is not a loss of support, such as gauge spread or track twist (e.g., recent incident at Grange-Over-Sands)
- Rail vehicle(s) roll-over due to overspeeding (e.g., Morpeth)
- Roll-over due to severe storm
The Trajectory Risk Model calculates the path of the derailed train for each derailment type – based on speed, curvature, and presence or otherwise of switches and crossings, but does not consider the effect that collision with structures, earthworks etc. has on the path of the derailed train.
Derailment mitigation will impact the trajectory.
The Escalation Model calculates the possible escalation of the derailment such as collision with structure, vehicles roll-over/fall, collision with another train, or fire/explosion. The event tree structure with probabilities is based on the Trajectory Risk Model.
And finally the Loss Model calculates the loss (safety/cost) for the base derailment plus any escalations from the Escalation Model.
So far, the principles of the model have been created with a small number of sample sections. Results for Carmont were illustrated for derailment risk, likely consequences and weighted fatalities index.
Next steps are to extend the model to the national network (approximately 600,000 25-metre sections) together with train types operating on each section, to incorporate Huddersfield’s work on the effectiveness of derailment mitigations (below) and the development of a simple user interface. Rail Engineer thinks this is a great deal of work to be delivered by the stated Autumn 2025.
The University of Huddersfield is simulating vehicle track interaction during the in-line phase of a derailment, post derailment containment including negative interactions e.g. with switches and crossing (which tend to make the outcome worse) and a parametric study on effectiveness of derailment containment for various operating conditions. The idea is that the model can be used to assess train or track features that might prevent trains deviating from plain line in the event of a derailment. But the first challenge is to build a modelling environment in which these features, or newly designed features can be evaluated.
Simpack-Rail was used to model up to the derailment point. This has the benefit that existing models could be used. Specific models were created in generic Simpack – the general-purpose multi-body dynamics package. Rails, guard rails, sleepers and ballast were modelled. The results of simulated trains running on sleepers/ballast were compared with published material on simulated and measured wheel response. The simulation had to consider wheel/ballast interaction including:
- Initial geometry – nominal profile, 3D geometry.
- Ballast surface deformation – non-linear force-deflection curve for a nominal wheel and inertial reaction from displaced ballast mass.
- Energy dissipation (longitudinal & lateral) – penetration depth and ballast characteristics and friction resistance from ballast displacement.
- Guidance effects – lateral reaction force (for the wheelset), inertial reaction force from ballast displacement, and friction sliding of wheel and ballast displacement.
- Parameters allow different conditions – soft ballast shoulder, compacted four-foot cribs.
It also allowed contact with submerged bodies/faces such as sleeper ends especially for duo block sleepers and there is provision for other surface types to be added later.
So far, the work has demonstrated that the simulation is feasible. The next steps move onto exploring post derailment mitigations, namely: (i) assessment of negative interaction of mitigations such as vehicle mounted and track mounted mitigations and vehicle mounted mitigations at switches and crossings; and (ii) application of the developed modelling capability into the derailment risk model.
The description above makes the modelling sound easy, but the speakers described their work as “pushing the bounds of modelling”.
RAIB Carmont Report recommendation 12:
“The intent of this recommendation is to take account of learning from the Carmont accident in the development of a coherent long-term strategy for derailment mitigation. It is anticipated that implementation of this recommendation will be informed by work, including RSSB project T1143, already undertaken by the rail industry as a result of recommendation 3 of RAIB’s investigation of the Watford derailment.
“RDG and Network Rail, in conjunction with RSSB, should consider and incorporate all relevant learning from the Carmont accident into the assessment of rolling stock and infrastructure design features that can provide guidance to trains when derailed. Particular features to be taken into account include:
- The risk of derailment from relatively small landslips and washouts.
- Position of track relative to adjacent ground on which derailed wheels may run (that is, features that can affect the deviation of a derailed train).
- Proximity to features with the potential to increase the consequence of an accident (bridge parapets, tunnel portals etc).
- Topography likely to increase the extent of vehicle scatter.
The above-mentioned assessment should then be used to develop a systemic, risk-based strategy for the provision of additional measures for the guidance of derailed trains that takes into account the appropriate balance between infrastructure-based mitigation and vehicle-based.”
Image credit: RAIB